Semi-structured information in XML can be merged in a logic-based framework [Hun02, Hun02b]. This framework has been extended to deal with uncertainty, in the form of probability values, degrees of beliefs, or necessity measures, associated with leaves (i.e., textentries) in the XML documents [HL04a]. In this paper we further extend this approach to modelling and merging uncertain information that is defined at different levels of granularity of XML textentries, and to modelling and reasoning with XML documents that contain semantically heterogeneous uncertain information on more complex elements in XML subtrees. We present the formal definitions for modelling, propagating and merging semantically heterogeneous uncertain information and explain how they can be handled using logic-based fusion techniques.